]> CyberLeo.Net >> Repos - FreeBSD/stable/10.git/blob - sys/netinet6/icmp6.c
MFC r275358 r275483 r276982 - Removing M_FLOWID by hps@
[FreeBSD/stable/10.git] / sys / netinet6 / icmp6.c
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *      $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30  */
31
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *      The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
61  */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69
70 #include <sys/param.h>
71 #include <sys/domain.h>
72 #include <sys/jail.h>
73 #include <sys/kernel.h>
74 #include <sys/lock.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/sx.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
85 #include <sys/time.h>
86
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/if_llatbl.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/vnet.h>
93
94 #include <netinet/in.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip6.h>
98 #include <netinet/icmp6.h>
99 #include <netinet/tcp_var.h>
100
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/ip6protosw.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/mld6_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/send.h>
109
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/key.h>
113 #endif
114
115 extern struct domain inet6domain;
116
117 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
118 VNET_PCPUSTAT_SYSINIT(icmp6stat);
119
120 #ifdef VIMAGE
121 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
122 #endif /* VIMAGE */
123
124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
125 VNET_DECLARE(struct inpcbhead, ripcb);
126 VNET_DECLARE(int, icmp6errppslim);
127 static VNET_DEFINE(int, icmp6errpps_count) = 0;
128 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
129 VNET_DECLARE(int, icmp6_nodeinfo);
130
131 #define V_ripcbinfo                     VNET(ripcbinfo)
132 #define V_ripcb                         VNET(ripcb)
133 #define V_icmp6errppslim                VNET(icmp6errppslim)
134 #define V_icmp6errpps_count             VNET(icmp6errpps_count)
135 #define V_icmp6errppslim_last           VNET(icmp6errppslim_last)
136 #define V_icmp6_nodeinfo                VNET(icmp6_nodeinfo)
137
138 static void icmp6_errcount(int, int);
139 static int icmp6_rip6_input(struct mbuf **, int);
140 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
141 static const char *icmp6_redirect_diag(struct in6_addr *,
142         struct in6_addr *, struct in6_addr *);
143 static struct mbuf *ni6_input(struct mbuf *, int);
144 static struct mbuf *ni6_nametodns(const char *, int, int);
145 static int ni6_dnsmatch(const char *, int, const char *, int);
146 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
147                           struct ifnet **, struct in6_addr *);
148 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
149                                 struct ifnet *, int);
150 static int icmp6_notify_error(struct mbuf **, int, int, int);
151
152 /*
153  * Kernel module interface for updating icmp6stat.  The argument is an index
154  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
155  * general layout of icmp6stat into the caller, it doesn't encode its
156  * location, so that future changes to add, for example, per-CPU stats
157  * support won't cause binary compatibility problems for kernel modules.
158  */
159 void
160 kmod_icmp6stat_inc(int statnum)
161 {
162
163         counter_u64_add(VNET(icmp6stat)[statnum], 1);
164 }
165
166 static void
167 icmp6_errcount(int type, int code)
168 {
169         switch (type) {
170         case ICMP6_DST_UNREACH:
171                 switch (code) {
172                 case ICMP6_DST_UNREACH_NOROUTE:
173                         ICMP6STAT_INC(icp6s_odst_unreach_noroute);
174                         return;
175                 case ICMP6_DST_UNREACH_ADMIN:
176                         ICMP6STAT_INC(icp6s_odst_unreach_admin);
177                         return;
178                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
179                         ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
180                         return;
181                 case ICMP6_DST_UNREACH_ADDR:
182                         ICMP6STAT_INC(icp6s_odst_unreach_addr);
183                         return;
184                 case ICMP6_DST_UNREACH_NOPORT:
185                         ICMP6STAT_INC(icp6s_odst_unreach_noport);
186                         return;
187                 }
188                 break;
189         case ICMP6_PACKET_TOO_BIG:
190                 ICMP6STAT_INC(icp6s_opacket_too_big);
191                 return;
192         case ICMP6_TIME_EXCEEDED:
193                 switch (code) {
194                 case ICMP6_TIME_EXCEED_TRANSIT:
195                         ICMP6STAT_INC(icp6s_otime_exceed_transit);
196                         return;
197                 case ICMP6_TIME_EXCEED_REASSEMBLY:
198                         ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
199                         return;
200                 }
201                 break;
202         case ICMP6_PARAM_PROB:
203                 switch (code) {
204                 case ICMP6_PARAMPROB_HEADER:
205                         ICMP6STAT_INC(icp6s_oparamprob_header);
206                         return;
207                 case ICMP6_PARAMPROB_NEXTHEADER:
208                         ICMP6STAT_INC(icp6s_oparamprob_nextheader);
209                         return;
210                 case ICMP6_PARAMPROB_OPTION:
211                         ICMP6STAT_INC(icp6s_oparamprob_option);
212                         return;
213                 }
214                 break;
215         case ND_REDIRECT:
216                 ICMP6STAT_INC(icp6s_oredirect);
217                 return;
218         }
219         ICMP6STAT_INC(icp6s_ounknown);
220 }
221
222 /*
223  * A wrapper function for icmp6_error() necessary when the erroneous packet
224  * may not contain enough scope zone information.
225  */
226 void
227 icmp6_error2(struct mbuf *m, int type, int code, int param,
228     struct ifnet *ifp)
229 {
230         struct ip6_hdr *ip6;
231
232         if (ifp == NULL)
233                 return;
234
235 #ifndef PULLDOWN_TEST
236         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
237 #else
238         if (m->m_len < sizeof(struct ip6_hdr)) {
239                 m = m_pullup(m, sizeof(struct ip6_hdr));
240                 if (m == NULL)
241                         return;
242         }
243 #endif
244
245         ip6 = mtod(m, struct ip6_hdr *);
246
247         if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
248                 return;
249         if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
250                 return;
251
252         icmp6_error(m, type, code, param);
253 }
254
255 /*
256  * Generate an error packet of type error in response to bad IP6 packet.
257  */
258 void
259 icmp6_error(struct mbuf *m, int type, int code, int param)
260 {
261         struct ip6_hdr *oip6, *nip6;
262         struct icmp6_hdr *icmp6;
263         u_int preplen;
264         int off;
265         int nxt;
266
267         ICMP6STAT_INC(icp6s_error);
268
269         /* count per-type-code statistics */
270         icmp6_errcount(type, code);
271
272 #ifdef M_DECRYPTED      /*not openbsd*/
273         if (m->m_flags & M_DECRYPTED) {
274                 ICMP6STAT_INC(icp6s_canterror);
275                 goto freeit;
276         }
277 #endif
278
279 #ifndef PULLDOWN_TEST
280         IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
281 #else
282         if (m->m_len < sizeof(struct ip6_hdr)) {
283                 m = m_pullup(m, sizeof(struct ip6_hdr));
284                 if (m == NULL)
285                         return;
286         }
287 #endif
288         oip6 = mtod(m, struct ip6_hdr *);
289
290         /*
291          * If the destination address of the erroneous packet is a multicast
292          * address, or the packet was sent using link-layer multicast,
293          * we should basically suppress sending an error (RFC 2463, Section
294          * 2.4).
295          * We have two exceptions (the item e.2 in that section):
296          * - the Packet Too Big message can be sent for path MTU discovery.
297          * - the Parameter Problem Message that can be allowed an icmp6 error
298          *   in the option type field.  This check has been done in
299          *   ip6_unknown_opt(), so we can just check the type and code.
300          */
301         if ((m->m_flags & (M_BCAST|M_MCAST) ||
302              IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
303             (type != ICMP6_PACKET_TOO_BIG &&
304              (type != ICMP6_PARAM_PROB ||
305               code != ICMP6_PARAMPROB_OPTION)))
306                 goto freeit;
307
308         /*
309          * RFC 2463, 2.4 (e.5): source address check.
310          * XXX: the case of anycast source?
311          */
312         if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
313             IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
314                 goto freeit;
315
316         /*
317          * If we are about to send ICMPv6 against ICMPv6 error/redirect,
318          * don't do it.
319          */
320         nxt = -1;
321         off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
322         if (off >= 0 && nxt == IPPROTO_ICMPV6) {
323                 struct icmp6_hdr *icp;
324
325 #ifndef PULLDOWN_TEST
326                 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
327                 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
328 #else
329                 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
330                         sizeof(*icp));
331                 if (icp == NULL) {
332                         ICMP6STAT_INC(icp6s_tooshort);
333                         return;
334                 }
335 #endif
336                 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
337                     icp->icmp6_type == ND_REDIRECT) {
338                         /*
339                          * ICMPv6 error
340                          * Special case: for redirect (which is
341                          * informational) we must not send icmp6 error.
342                          */
343                         ICMP6STAT_INC(icp6s_canterror);
344                         goto freeit;
345                 } else {
346                         /* ICMPv6 informational - send the error */
347                 }
348         } else {
349                 /* non-ICMPv6 - send the error */
350         }
351
352         oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
353
354         /* Finally, do rate limitation check. */
355         if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
356                 ICMP6STAT_INC(icp6s_toofreq);
357                 goto freeit;
358         }
359
360         /*
361          * OK, ICMP6 can be generated.
362          */
363
364         if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
365                 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
366
367         preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
368         M_PREPEND(m, preplen, M_NOWAIT);        /* FIB is also copied over. */
369         if (m == NULL) {
370                 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
371                 return;
372         }
373
374         nip6 = mtod(m, struct ip6_hdr *);
375         nip6->ip6_src  = oip6->ip6_src;
376         nip6->ip6_dst  = oip6->ip6_dst;
377
378         in6_clearscope(&oip6->ip6_src);
379         in6_clearscope(&oip6->ip6_dst);
380
381         icmp6 = (struct icmp6_hdr *)(nip6 + 1);
382         icmp6->icmp6_type = type;
383         icmp6->icmp6_code = code;
384         icmp6->icmp6_pptr = htonl((u_int32_t)param);
385
386         /*
387          * icmp6_reflect() is designed to be in the input path.
388          * icmp6_error() can be called from both input and output path,
389          * and if we are in output path rcvif could contain bogus value.
390          * clear m->m_pkthdr.rcvif for safety, we should have enough scope
391          * information in ip header (nip6).
392          */
393         m->m_pkthdr.rcvif = NULL;
394
395         ICMP6STAT_INC(icp6s_outhist[type]);
396         icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
397
398         return;
399
400   freeit:
401         /*
402          * If we can't tell whether or not we can generate ICMP6, free it.
403          */
404         m_freem(m);
405 }
406
407 /*
408  * Process a received ICMP6 message.
409  */
410 int
411 icmp6_input(struct mbuf **mp, int *offp, int proto)
412 {
413         struct mbuf *m = *mp, *n;
414         struct ifnet *ifp;
415         struct ip6_hdr *ip6, *nip6;
416         struct icmp6_hdr *icmp6, *nicmp6;
417         int off = *offp;
418         int icmp6len = m->m_pkthdr.len - *offp;
419         int code, sum, noff;
420         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
421         int ip6len, error;
422
423         ifp = m->m_pkthdr.rcvif;
424
425 #ifndef PULLDOWN_TEST
426         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
427         /* m might change if M_LOOP.  So, call mtod after this */
428 #endif
429
430         /*
431          * Locate icmp6 structure in mbuf, and check
432          * that not corrupted and of at least minimum length
433          */
434
435         ip6 = mtod(m, struct ip6_hdr *);
436         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
437         if (icmp6len < sizeof(struct icmp6_hdr)) {
438                 ICMP6STAT_INC(icp6s_tooshort);
439                 goto freeit;
440         }
441
442         /*
443          * Check multicast group membership.
444          * Note: SSM filters are not applied for ICMPv6 traffic.
445          */
446         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
447                 struct in6_multi        *inm;
448
449                 inm = in6m_lookup(ifp, &ip6->ip6_dst);
450                 if (inm == NULL) {
451                         IP6STAT_INC(ip6s_notmember);
452                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
453                         goto freeit;
454                 }
455         }
456
457         /*
458          * calculate the checksum
459          */
460 #ifndef PULLDOWN_TEST
461         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
462 #else
463         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
464         if (icmp6 == NULL) {
465                 ICMP6STAT_INC(icp6s_tooshort);
466                 return IPPROTO_DONE;
467         }
468 #endif
469         code = icmp6->icmp6_code;
470
471         if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
472                 nd6log((LOG_ERR,
473                     "ICMP6 checksum error(%d|%x) %s\n",
474                     icmp6->icmp6_type, sum,
475                     ip6_sprintf(ip6bufs, &ip6->ip6_src)));
476                 ICMP6STAT_INC(icp6s_checksum);
477                 goto freeit;
478         }
479
480         if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
481                 /*
482                  * Deliver very specific ICMP6 type only.
483                  * This is important to deliver TOOBIG.  Otherwise PMTUD
484                  * will not work.
485                  */
486                 switch (icmp6->icmp6_type) {
487                 case ICMP6_DST_UNREACH:
488                 case ICMP6_PACKET_TOO_BIG:
489                 case ICMP6_TIME_EXCEEDED:
490                         break;
491                 default:
492                         goto freeit;
493                 }
494         }
495
496         ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
497         icmp6_ifstat_inc(ifp, ifs6_in_msg);
498         if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
499                 icmp6_ifstat_inc(ifp, ifs6_in_error);
500
501         switch (icmp6->icmp6_type) {
502         case ICMP6_DST_UNREACH:
503                 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
504                 switch (code) {
505                 case ICMP6_DST_UNREACH_NOROUTE:
506                         code = PRC_UNREACH_NET;
507                         break;
508                 case ICMP6_DST_UNREACH_ADMIN:
509                         icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
510                         code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
511                         break;
512                 case ICMP6_DST_UNREACH_ADDR:
513                         code = PRC_HOSTDEAD;
514                         break;
515                 case ICMP6_DST_UNREACH_BEYONDSCOPE:
516                         /* I mean "source address was incorrect." */
517                         code = PRC_PARAMPROB;
518                         break;
519                 case ICMP6_DST_UNREACH_NOPORT:
520                         code = PRC_UNREACH_PORT;
521                         break;
522                 default:
523                         goto badcode;
524                 }
525                 goto deliver;
526                 break;
527
528         case ICMP6_PACKET_TOO_BIG:
529                 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
530
531                 /* validation is made in icmp6_mtudisc_update */
532
533                 code = PRC_MSGSIZE;
534
535                 /*
536                  * Updating the path MTU will be done after examining
537                  * intermediate extension headers.
538                  */
539                 goto deliver;
540                 break;
541
542         case ICMP6_TIME_EXCEEDED:
543                 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
544                 switch (code) {
545                 case ICMP6_TIME_EXCEED_TRANSIT:
546                         code = PRC_TIMXCEED_INTRANS;
547                         break;
548                 case ICMP6_TIME_EXCEED_REASSEMBLY:
549                         code = PRC_TIMXCEED_REASS;
550                         break;
551                 default:
552                         goto badcode;
553                 }
554                 goto deliver;
555                 break;
556
557         case ICMP6_PARAM_PROB:
558                 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
559                 switch (code) {
560                 case ICMP6_PARAMPROB_NEXTHEADER:
561                         code = PRC_UNREACH_PROTOCOL;
562                         break;
563                 case ICMP6_PARAMPROB_HEADER:
564                 case ICMP6_PARAMPROB_OPTION:
565                         code = PRC_PARAMPROB;
566                         break;
567                 default:
568                         goto badcode;
569                 }
570                 goto deliver;
571                 break;
572
573         case ICMP6_ECHO_REQUEST:
574                 icmp6_ifstat_inc(ifp, ifs6_in_echo);
575                 if (code != 0)
576                         goto badcode;
577                 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
578                         /* Give up remote */
579                         break;
580                 }
581                 if ((n->m_flags & M_EXT) != 0
582                  || n->m_len < off + sizeof(struct icmp6_hdr)) {
583                         struct mbuf *n0 = n;
584                         int n0len;
585
586                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
587                         n = m_gethdr(M_NOWAIT, n0->m_type);
588                         if (n == NULL) {
589                                 /* Give up remote */
590                                 m_freem(n0);
591                                 break;
592                         }
593
594                         m_move_pkthdr(n, n0);   /* FIB copied. */
595                         n0len = n0->m_pkthdr.len;       /* save for use below */
596                         /*
597                          * Copy IPv6 and ICMPv6 only.
598                          */
599                         nip6 = mtod(n, struct ip6_hdr *);
600                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
601                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
602                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
603                         noff = sizeof(struct ip6_hdr);
604                         /* new mbuf contains only ipv6+icmpv6 headers */
605                         n->m_len = noff + sizeof(struct icmp6_hdr);
606                         /*
607                          * Adjust mbuf.  ip6_plen will be adjusted in
608                          * ip6_output().
609                          */
610                         m_adj(n0, off + sizeof(struct icmp6_hdr));
611                         /* recalculate complete packet size */
612                         n->m_pkthdr.len = n0len + (noff - off);
613                         n->m_next = n0;
614                 } else {
615                         nip6 = mtod(n, struct ip6_hdr *);
616                         IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
617                             sizeof(*nicmp6));
618                         noff = off;
619                 }
620                 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
621                 nicmp6->icmp6_code = 0;
622                 if (n) {
623                         ICMP6STAT_INC(icp6s_reflect);
624                         ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
625                         icmp6_reflect(n, noff);
626                 }
627                 break;
628
629         case ICMP6_ECHO_REPLY:
630                 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
631                 if (code != 0)
632                         goto badcode;
633                 break;
634
635         case MLD_LISTENER_QUERY:
636         case MLD_LISTENER_REPORT:
637         case MLD_LISTENER_DONE:
638         case MLDV2_LISTENER_REPORT:
639                 /*
640                  * Drop MLD traffic which is not link-local, has a hop limit
641                  * of greater than 1 hop, or which does not have the
642                  * IPv6 HBH Router Alert option.
643                  * As IPv6 HBH options are stripped in ip6_input() we must
644                  * check an mbuf header flag.
645                  * XXX Should we also sanity check that these messages
646                  * were directed to a link-local multicast prefix?
647                  */
648                 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
649                         goto freeit;
650                 if (mld_input(m, off, icmp6len) != 0)
651                         return (IPPROTO_DONE);
652                 /* m stays. */
653                 break;
654
655         case ICMP6_WRUREQUEST:  /* ICMP6_FQDN_QUERY */
656             {
657                 enum { WRU, FQDN } mode;
658
659                 if (!V_icmp6_nodeinfo)
660                         break;
661
662                 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
663                         mode = WRU;
664                 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
665                         mode = FQDN;
666                 else
667                         goto badlen;
668
669                 if (mode == FQDN) {
670 #ifndef PULLDOWN_TEST
671                         IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
672                             IPPROTO_DONE);
673 #endif
674                         n = m_copy(m, 0, M_COPYALL);
675                         if (n)
676                                 n = ni6_input(n, off);
677                         /* XXX meaningless if n == NULL */
678                         noff = sizeof(struct ip6_hdr);
679                 } else {
680                         struct prison *pr;
681                         u_char *p;
682                         int maxhlen, hlen;
683
684                         /*
685                          * XXX: this combination of flags is pointless,
686                          * but should we keep this for compatibility?
687                          */
688                         if ((V_icmp6_nodeinfo & 5) != 5)
689                                 break;
690
691                         if (code != 0)
692                                 goto badcode;
693
694                         CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
695                         n = m_gethdr(M_NOWAIT, m->m_type);
696                         if (n == NULL) {
697                                 /* Give up remote */
698                                 break;
699                         }
700                         if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
701                                 /*
702                                  * Previous code did a blind M_COPY_PKTHDR
703                                  * and said "just for rcvif".  If true, then
704                                  * we could tolerate the dup failing (due to
705                                  * the deep copy of the tag chain).  For now
706                                  * be conservative and just fail.
707                                  */
708                                 m_free(n);
709                                 n = NULL;
710                         }
711                         maxhlen = M_TRAILINGSPACE(n) -
712                             (sizeof(*nip6) + sizeof(*nicmp6) + 4);
713                         pr = curthread->td_ucred->cr_prison;
714                         mtx_lock(&pr->pr_mtx);
715                         hlen = strlen(pr->pr_hostname);
716                         if (maxhlen > hlen)
717                                 maxhlen = hlen;
718                         /*
719                          * Copy IPv6 and ICMPv6 only.
720                          */
721                         nip6 = mtod(n, struct ip6_hdr *);
722                         bcopy(ip6, nip6, sizeof(struct ip6_hdr));
723                         nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
724                         bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
725                         p = (u_char *)(nicmp6 + 1);
726                         bzero(p, 4);
727                         /* meaningless TTL */
728                         bcopy(pr->pr_hostname, p + 4, maxhlen);
729                         mtx_unlock(&pr->pr_mtx);
730                         noff = sizeof(struct ip6_hdr);
731                         n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
732                                 sizeof(struct icmp6_hdr) + 4 + maxhlen;
733                         nicmp6->icmp6_type = ICMP6_WRUREPLY;
734                         nicmp6->icmp6_code = 0;
735                 }
736                 if (n) {
737                         ICMP6STAT_INC(icp6s_reflect);
738                         ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
739                         icmp6_reflect(n, noff);
740                 }
741                 break;
742             }
743
744         case ICMP6_WRUREPLY:
745                 if (code != 0)
746                         goto badcode;
747                 break;
748
749         case ND_ROUTER_SOLICIT:
750                 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
751                 if (code != 0)
752                         goto badcode;
753                 if (icmp6len < sizeof(struct nd_router_solicit))
754                         goto badlen;
755                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
756                         /* give up local */
757
758                         /* Send incoming SeND packet to user space. */
759                         if (send_sendso_input_hook != NULL) {
760                                 IP6_EXTHDR_CHECK(m, off,
761                                     icmp6len, IPPROTO_DONE);
762                                 error = send_sendso_input_hook(m, ifp,
763                                     SND_IN, ip6len);
764                                 /* -1 == no app on SEND socket */
765                                 if (error == 0)
766                                         return (IPPROTO_DONE);
767                                 nd6_rs_input(m, off, icmp6len);
768                         } else
769                                 nd6_rs_input(m, off, icmp6len);
770                         m = NULL;
771                         goto freeit;
772                 }
773                 if (send_sendso_input_hook != NULL) {
774                         IP6_EXTHDR_CHECK(n, off,
775                             icmp6len, IPPROTO_DONE);
776                         error = send_sendso_input_hook(n, ifp,
777                             SND_IN, ip6len);
778                         if (error == 0)
779                                 goto freeit;
780                         /* -1 == no app on SEND socket */
781                         nd6_rs_input(n, off, icmp6len);
782                 } else
783                         nd6_rs_input(n, off, icmp6len);
784                 /* m stays. */
785                 break;
786
787         case ND_ROUTER_ADVERT:
788                 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
789                 if (code != 0)
790                         goto badcode;
791                 if (icmp6len < sizeof(struct nd_router_advert))
792                         goto badlen;
793                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
794
795                         /* Send incoming SeND-protected/ND packet to user space. */
796                         if (send_sendso_input_hook != NULL) {
797                                 error = send_sendso_input_hook(m, ifp,
798                                     SND_IN, ip6len);
799                                 if (error == 0)
800                                         return (IPPROTO_DONE);
801                                 nd6_ra_input(m, off, icmp6len);
802                         } else
803                                 nd6_ra_input(m, off, icmp6len);
804                         m = NULL;
805                         goto freeit;
806                 }
807                 if (send_sendso_input_hook != NULL) {
808                         error = send_sendso_input_hook(n, ifp,
809                             SND_IN, ip6len);
810                         if (error == 0)
811                                 goto freeit;
812                         nd6_ra_input(n, off, icmp6len);
813                 } else
814                         nd6_ra_input(n, off, icmp6len);
815                 /* m stays. */
816                 break;
817
818         case ND_NEIGHBOR_SOLICIT:
819                 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
820                 if (code != 0)
821                         goto badcode;
822                 if (icmp6len < sizeof(struct nd_neighbor_solicit))
823                         goto badlen;
824                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
825                         if (send_sendso_input_hook != NULL) {
826                                 error = send_sendso_input_hook(m, ifp,
827                                     SND_IN, ip6len);
828                                 if (error == 0)
829                                         return (IPPROTO_DONE);
830                                 nd6_ns_input(m, off, icmp6len);
831                         } else
832                                 nd6_ns_input(m, off, icmp6len);
833                         m = NULL;
834                         goto freeit;
835                 }
836                 if (send_sendso_input_hook != NULL) {
837                         error = send_sendso_input_hook(n, ifp,
838                             SND_IN, ip6len);
839                         if (error == 0)
840                                 goto freeit;
841                         nd6_ns_input(n, off, icmp6len);
842                 } else
843                         nd6_ns_input(n, off, icmp6len);
844                 /* m stays. */
845                 break;
846
847         case ND_NEIGHBOR_ADVERT:
848                 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
849                 if (code != 0)
850                         goto badcode;
851                 if (icmp6len < sizeof(struct nd_neighbor_advert))
852                         goto badlen;
853                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
854
855                         /* Send incoming SeND-protected/ND packet to user space. */
856                         if (send_sendso_input_hook != NULL) {
857                                 error = send_sendso_input_hook(m, ifp,
858                                     SND_IN, ip6len);
859                                 if (error == 0)
860                                         return (IPPROTO_DONE);
861                                 nd6_na_input(m, off, icmp6len);
862                         } else
863                                 nd6_na_input(m, off, icmp6len);
864                         m = NULL;
865                         goto freeit;
866                 }
867                 if (send_sendso_input_hook != NULL) {
868                         error = send_sendso_input_hook(n, ifp,
869                             SND_IN, ip6len);
870                         if (error == 0)
871                                 goto freeit;
872                         nd6_na_input(n, off, icmp6len);
873                 } else
874                         nd6_na_input(n, off, icmp6len);
875                 /* m stays. */
876                 break;
877
878         case ND_REDIRECT:
879                 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
880                 if (code != 0)
881                         goto badcode;
882                 if (icmp6len < sizeof(struct nd_redirect))
883                         goto badlen;
884                 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
885                         if (send_sendso_input_hook != NULL) {
886                                 error = send_sendso_input_hook(m, ifp,
887                                     SND_IN, ip6len);
888                                 if (error == 0)
889                                         return (IPPROTO_DONE);
890                             icmp6_redirect_input(m, off);
891                         } else
892                                 icmp6_redirect_input(m, off);
893                         m = NULL;
894                         goto freeit;
895                 }
896                 if (send_sendso_input_hook != NULL) {
897                         error = send_sendso_input_hook(n, ifp,
898                             SND_IN, ip6len);
899                         if (error == 0)
900                                 goto freeit;
901                         icmp6_redirect_input(n, off);
902                 } else
903                         icmp6_redirect_input(n, off);
904                 /* m stays. */
905                 break;
906
907         case ICMP6_ROUTER_RENUMBERING:
908                 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
909                     code != ICMP6_ROUTER_RENUMBERING_RESULT)
910                         goto badcode;
911                 if (icmp6len < sizeof(struct icmp6_router_renum))
912                         goto badlen;
913                 break;
914
915         default:
916                 nd6log((LOG_DEBUG,
917                     "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
918                     icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
919                     ip6_sprintf(ip6bufd, &ip6->ip6_dst),
920                     ifp ? ifp->if_index : 0));
921                 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
922                         /* ICMPv6 error: MUST deliver it by spec... */
923                         code = PRC_NCMDS;
924                         /* deliver */
925                 } else {
926                         /* ICMPv6 informational: MUST not deliver */
927                         break;
928                 }
929         deliver:
930                 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
931                         /* In this case, m should've been freed. */
932                         return (IPPROTO_DONE);
933                 }
934                 break;
935
936         badcode:
937                 ICMP6STAT_INC(icp6s_badcode);
938                 break;
939
940         badlen:
941                 ICMP6STAT_INC(icp6s_badlen);
942                 break;
943         }
944
945         /* deliver the packet to appropriate sockets */
946         icmp6_rip6_input(&m, *offp);
947
948         return IPPROTO_DONE;
949
950  freeit:
951         m_freem(m);
952         return IPPROTO_DONE;
953 }
954
955 static int
956 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
957 {
958         struct mbuf *m = *mp;
959         struct icmp6_hdr *icmp6;
960         struct ip6_hdr *eip6;
961         u_int32_t notifymtu;
962         struct sockaddr_in6 icmp6src, icmp6dst;
963
964         if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
965                 ICMP6STAT_INC(icp6s_tooshort);
966                 goto freeit;
967         }
968 #ifndef PULLDOWN_TEST
969         IP6_EXTHDR_CHECK(m, off,
970             sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
971         icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
972 #else
973         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
974             sizeof(*icmp6) + sizeof(struct ip6_hdr));
975         if (icmp6 == NULL) {
976                 ICMP6STAT_INC(icp6s_tooshort);
977                 return (-1);
978         }
979 #endif
980         eip6 = (struct ip6_hdr *)(icmp6 + 1);
981
982         /* Detect the upper level protocol */
983         {
984                 void (*ctlfunc)(int, struct sockaddr *, void *);
985                 u_int8_t nxt = eip6->ip6_nxt;
986                 int eoff = off + sizeof(struct icmp6_hdr) +
987                     sizeof(struct ip6_hdr);
988                 struct ip6ctlparam ip6cp;
989                 struct in6_addr *finaldst = NULL;
990                 int icmp6type = icmp6->icmp6_type;
991                 struct ip6_frag *fh;
992                 struct ip6_rthdr *rth;
993                 struct ip6_rthdr0 *rth0;
994                 int rthlen;
995
996                 while (1) { /* XXX: should avoid infinite loop explicitly? */
997                         struct ip6_ext *eh;
998
999                         switch (nxt) {
1000                         case IPPROTO_HOPOPTS:
1001                         case IPPROTO_DSTOPTS:
1002                         case IPPROTO_AH:
1003 #ifndef PULLDOWN_TEST
1004                                 IP6_EXTHDR_CHECK(m, 0,
1005                                     eoff + sizeof(struct ip6_ext), -1);
1006                                 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
1007 #else
1008                                 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1009                                     eoff, sizeof(*eh));
1010                                 if (eh == NULL) {
1011                                         ICMP6STAT_INC(icp6s_tooshort);
1012                                         return (-1);
1013                                 }
1014 #endif
1015
1016                                 if (nxt == IPPROTO_AH)
1017                                         eoff += (eh->ip6e_len + 2) << 2;
1018                                 else
1019                                         eoff += (eh->ip6e_len + 1) << 3;
1020                                 nxt = eh->ip6e_nxt;
1021                                 break;
1022                         case IPPROTO_ROUTING:
1023                                 /*
1024                                  * When the erroneous packet contains a
1025                                  * routing header, we should examine the
1026                                  * header to determine the final destination.
1027                                  * Otherwise, we can't properly update
1028                                  * information that depends on the final
1029                                  * destination (e.g. path MTU).
1030                                  */
1031 #ifndef PULLDOWN_TEST
1032                                 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
1033                                 rth = (struct ip6_rthdr *)
1034                                     (mtod(m, caddr_t) + eoff);
1035 #else
1036                                 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1037                                     eoff, sizeof(*rth));
1038                                 if (rth == NULL) {
1039                                         ICMP6STAT_INC(icp6s_tooshort);
1040                                         return (-1);
1041                                 }
1042 #endif
1043                                 rthlen = (rth->ip6r_len + 1) << 3;
1044                                 /*
1045                                  * XXX: currently there is no
1046                                  * officially defined type other
1047                                  * than type-0.
1048                                  * Note that if the segment left field
1049                                  * is 0, all intermediate hops must
1050                                  * have been passed.
1051                                  */
1052                                 if (rth->ip6r_segleft &&
1053                                     rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1054                                         int hops;
1055
1056 #ifndef PULLDOWN_TEST
1057                                         IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1058                                         rth0 = (struct ip6_rthdr0 *)
1059                                             (mtod(m, caddr_t) + eoff);
1060 #else
1061                                         IP6_EXTHDR_GET(rth0,
1062                                             struct ip6_rthdr0 *, m,
1063                                             eoff, rthlen);
1064                                         if (rth0 == NULL) {
1065                                                 ICMP6STAT_INC(icp6s_tooshort);
1066                                                 return (-1);
1067                                         }
1068 #endif
1069                                         /* just ignore a bogus header */
1070                                         if ((rth0->ip6r0_len % 2) == 0 &&
1071                                             (hops = rth0->ip6r0_len/2))
1072                                                 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1073                                 }
1074                                 eoff += rthlen;
1075                                 nxt = rth->ip6r_nxt;
1076                                 break;
1077                         case IPPROTO_FRAGMENT:
1078 #ifndef PULLDOWN_TEST
1079                                 IP6_EXTHDR_CHECK(m, 0, eoff +
1080                                     sizeof(struct ip6_frag), -1);
1081                                 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1082                                     eoff);
1083 #else
1084                                 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1085                                     eoff, sizeof(*fh));
1086                                 if (fh == NULL) {
1087                                         ICMP6STAT_INC(icp6s_tooshort);
1088                                         return (-1);
1089                                 }
1090 #endif
1091                                 /*
1092                                  * Data after a fragment header is meaningless
1093                                  * unless it is the first fragment, but
1094                                  * we'll go to the notify label for path MTU
1095                                  * discovery.
1096                                  */
1097                                 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1098                                         goto notify;
1099
1100                                 eoff += sizeof(struct ip6_frag);
1101                                 nxt = fh->ip6f_nxt;
1102                                 break;
1103                         default:
1104                                 /*
1105                                  * This case includes ESP and the No Next
1106                                  * Header.  In such cases going to the notify
1107                                  * label does not have any meaning
1108                                  * (i.e. ctlfunc will be NULL), but we go
1109                                  * anyway since we might have to update
1110                                  * path MTU information.
1111                                  */
1112                                 goto notify;
1113                         }
1114                 }
1115           notify:
1116 #ifndef PULLDOWN_TEST
1117                 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1118 #else
1119                 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1120                     sizeof(*icmp6) + sizeof(struct ip6_hdr));
1121                 if (icmp6 == NULL) {
1122                         ICMP6STAT_INC(icp6s_tooshort);
1123                         return (-1);
1124                 }
1125 #endif
1126
1127                 /*
1128                  * retrieve parameters from the inner IPv6 header, and convert
1129                  * them into sockaddr structures.
1130                  * XXX: there is no guarantee that the source or destination
1131                  * addresses of the inner packet are in the same scope as
1132                  * the addresses of the icmp packet.  But there is no other
1133                  * way to determine the zone.
1134                  */
1135                 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1136
1137                 bzero(&icmp6dst, sizeof(icmp6dst));
1138                 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1139                 icmp6dst.sin6_family = AF_INET6;
1140                 if (finaldst == NULL)
1141                         icmp6dst.sin6_addr = eip6->ip6_dst;
1142                 else
1143                         icmp6dst.sin6_addr = *finaldst;
1144                 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1145                         goto freeit;
1146                 bzero(&icmp6src, sizeof(icmp6src));
1147                 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1148                 icmp6src.sin6_family = AF_INET6;
1149                 icmp6src.sin6_addr = eip6->ip6_src;
1150                 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1151                         goto freeit;
1152                 icmp6src.sin6_flowinfo =
1153                     (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1154
1155                 if (finaldst == NULL)
1156                         finaldst = &eip6->ip6_dst;
1157                 ip6cp.ip6c_m = m;
1158                 ip6cp.ip6c_icmp6 = icmp6;
1159                 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1160                 ip6cp.ip6c_off = eoff;
1161                 ip6cp.ip6c_finaldst = finaldst;
1162                 ip6cp.ip6c_src = &icmp6src;
1163                 ip6cp.ip6c_nxt = nxt;
1164
1165                 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1166                         notifymtu = ntohl(icmp6->icmp6_mtu);
1167                         ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1168                         icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1169                 }
1170
1171                 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1172                     (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1173                 if (ctlfunc) {
1174                         (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1175                             &ip6cp);
1176                 }
1177         }
1178         *mp = m;
1179         return (0);
1180
1181   freeit:
1182         m_freem(m);
1183         return (-1);
1184 }
1185
1186 void
1187 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1188 {
1189         struct in6_addr *dst = ip6cp->ip6c_finaldst;
1190         struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1191         struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1192         u_int mtu = ntohl(icmp6->icmp6_mtu);
1193         struct in_conninfo inc;
1194
1195 #if 0
1196         /*
1197          * RFC2460 section 5, last paragraph.
1198          * even though minimum link MTU for IPv6 is IPV6_MMTU,
1199          * we may see ICMPv6 too big with mtu < IPV6_MMTU
1200          * due to packet translator in the middle.
1201          * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1202          * special handling.
1203          */
1204         if (mtu < IPV6_MMTU)
1205                 return;
1206 #endif
1207
1208         /*
1209          * we reject ICMPv6 too big with abnormally small value.
1210          * XXX what is the good definition of "abnormally small"?
1211          */
1212         if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1213                 return;
1214
1215         if (!validated)
1216                 return;
1217
1218         /*
1219          * In case the suggested mtu is less than IPV6_MMTU, we
1220          * only need to remember that it was for above mentioned
1221          * "alwaysfrag" case.
1222          * Try to be as close to the spec as possible.
1223          */
1224         if (mtu < IPV6_MMTU)
1225                 mtu = IPV6_MMTU - 8;
1226
1227         bzero(&inc, sizeof(inc));
1228         inc.inc_fibnum = M_GETFIB(m);
1229         inc.inc_flags |= INC_ISIPV6;
1230         inc.inc6_faddr = *dst;
1231         if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1232                 return;
1233
1234         if (mtu < tcp_maxmtu6(&inc, NULL)) {
1235                 tcp_hc_updatemtu(&inc, mtu);
1236                 ICMP6STAT_INC(icp6s_pmtuchg);
1237         }
1238 }
1239
1240 /*
1241  * Process a Node Information Query packet, based on
1242  * draft-ietf-ipngwg-icmp-name-lookups-07.
1243  *
1244  * Spec incompatibilities:
1245  * - IPv6 Subject address handling
1246  * - IPv4 Subject address handling support missing
1247  * - Proxy reply (answer even if it's not for me)
1248  * - joins NI group address at in6_ifattach() time only, does not cope
1249  *   with hostname changes by sethostname(3)
1250  */
1251 static struct mbuf *
1252 ni6_input(struct mbuf *m, int off)
1253 {
1254         struct icmp6_nodeinfo *ni6, *nni6;
1255         struct mbuf *n = NULL;
1256         struct prison *pr;
1257         u_int16_t qtype;
1258         int subjlen;
1259         int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1260         struct ni_reply_fqdn *fqdn;
1261         int addrs;              /* for NI_QTYPE_NODEADDR */
1262         struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1263         struct in6_addr in6_subj; /* subject address */
1264         struct ip6_hdr *ip6;
1265         int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1266         char *subj = NULL;
1267         struct in6_ifaddr *ia6 = NULL;
1268
1269         ip6 = mtod(m, struct ip6_hdr *);
1270 #ifndef PULLDOWN_TEST
1271         ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1272 #else
1273         IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1274         if (ni6 == NULL) {
1275                 /* m is already reclaimed */
1276                 return (NULL);
1277         }
1278 #endif
1279
1280         /*
1281          * Validate IPv6 source address.
1282          * The default configuration MUST be to refuse answering queries from
1283          * global-scope addresses according to RFC4602.
1284          * Notes:
1285          *  - it's not very clear what "refuse" means; this implementation
1286          *    simply drops it.
1287          *  - it's not very easy to identify global-scope (unicast) addresses
1288          *    since there are many prefixes for them.  It should be safer
1289          *    and in practice sufficient to check "all" but loopback and
1290          *    link-local (note that site-local unicast was deprecated and
1291          *    ULA is defined as global scope-wise)
1292          */
1293         if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1294             !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1295             !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1296                 goto bad;
1297
1298         /*
1299          * Validate IPv6 destination address.
1300          *
1301          * The Responder must discard the Query without further processing
1302          * unless it is one of the Responder's unicast or anycast addresses, or
1303          * a link-local scope multicast address which the Responder has joined.
1304          * [RFC4602, Section 5.]
1305          */
1306         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1307                 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1308                         goto bad;
1309                 /* else it's a link-local multicast, fine */
1310         } else {                /* unicast or anycast */
1311                 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1312                         goto bad; /* XXX impossible */
1313
1314                 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1315                     !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1316                         ifa_free(&ia6->ia_ifa);
1317                         nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1318                                 "a temporary address in %s:%d",
1319                                __FILE__, __LINE__));
1320                         goto bad;
1321                 }
1322                 ifa_free(&ia6->ia_ifa);
1323         }
1324
1325         /* validate query Subject field. */
1326         qtype = ntohs(ni6->ni_qtype);
1327         subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1328         switch (qtype) {
1329         case NI_QTYPE_NOOP:
1330         case NI_QTYPE_SUPTYPES:
1331                 /* 07 draft */
1332                 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1333                         break;
1334                 /* FALLTHROUGH */
1335         case NI_QTYPE_FQDN:
1336         case NI_QTYPE_NODEADDR:
1337         case NI_QTYPE_IPV4ADDR:
1338                 switch (ni6->ni_code) {
1339                 case ICMP6_NI_SUBJ_IPV6:
1340 #if ICMP6_NI_SUBJ_IPV6 != 0
1341                 case 0:
1342 #endif
1343                         /*
1344                          * backward compatibility - try to accept 03 draft
1345                          * format, where no Subject is present.
1346                          */
1347                         if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1348                             subjlen == 0) {
1349                                 oldfqdn++;
1350                                 break;
1351                         }
1352 #if ICMP6_NI_SUBJ_IPV6 != 0
1353                         if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1354                                 goto bad;
1355 #endif
1356
1357                         if (subjlen != sizeof(struct in6_addr))
1358                                 goto bad;
1359
1360                         /*
1361                          * Validate Subject address.
1362                          *
1363                          * Not sure what exactly "address belongs to the node"
1364                          * means in the spec, is it just unicast, or what?
1365                          *
1366                          * At this moment we consider Subject address as
1367                          * "belong to the node" if the Subject address equals
1368                          * to the IPv6 destination address; validation for
1369                          * IPv6 destination address should have done enough
1370                          * check for us.
1371                          *
1372                          * We do not do proxy at this moment.
1373                          */
1374                         /* m_pulldown instead of copy? */
1375                         m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1376                             subjlen, (caddr_t)&in6_subj);
1377                         if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1378                                 goto bad;
1379
1380                         subj = (char *)&in6_subj;
1381                         if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1382                                 break;
1383
1384                         /*
1385                          * XXX if we are to allow other cases, we should really
1386                          * be careful about scope here.
1387                          * basically, we should disallow queries toward IPv6
1388                          * destination X with subject Y,
1389                          * if scope(X) > scope(Y).
1390                          * if we allow scope(X) > scope(Y), it will result in
1391                          * information leakage across scope boundary.
1392                          */
1393                         goto bad;
1394
1395                 case ICMP6_NI_SUBJ_FQDN:
1396                         /*
1397                          * Validate Subject name with gethostname(3).
1398                          *
1399                          * The behavior may need some debate, since:
1400                          * - we are not sure if the node has FQDN as
1401                          *   hostname (returned by gethostname(3)).
1402                          * - the code does wildcard match for truncated names.
1403                          *   however, we are not sure if we want to perform
1404                          *   wildcard match, if gethostname(3) side has
1405                          *   truncated hostname.
1406                          */
1407                         pr = curthread->td_ucred->cr_prison;
1408                         mtx_lock(&pr->pr_mtx);
1409                         n = ni6_nametodns(pr->pr_hostname,
1410                             strlen(pr->pr_hostname), 0);
1411                         mtx_unlock(&pr->pr_mtx);
1412                         if (!n || n->m_next || n->m_len == 0)
1413                                 goto bad;
1414                         IP6_EXTHDR_GET(subj, char *, m,
1415                             off + sizeof(struct icmp6_nodeinfo), subjlen);
1416                         if (subj == NULL)
1417                                 goto bad;
1418                         if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1419                             n->m_len)) {
1420                                 goto bad;
1421                         }
1422                         m_freem(n);
1423                         n = NULL;
1424                         break;
1425
1426                 case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1427                 default:
1428                         goto bad;
1429                 }
1430                 break;
1431         }
1432
1433         /* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1434         switch (qtype) {
1435         case NI_QTYPE_FQDN:
1436                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1437                         goto bad;
1438                 break;
1439         case NI_QTYPE_NODEADDR:
1440         case NI_QTYPE_IPV4ADDR:
1441                 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1442                         goto bad;
1443                 break;
1444         }
1445
1446         /* guess reply length */
1447         switch (qtype) {
1448         case NI_QTYPE_NOOP:
1449                 break;          /* no reply data */
1450         case NI_QTYPE_SUPTYPES:
1451                 replylen += sizeof(u_int32_t);
1452                 break;
1453         case NI_QTYPE_FQDN:
1454                 /* XXX will append an mbuf */
1455                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1456                 break;
1457         case NI_QTYPE_NODEADDR:
1458                 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1459                 if ((replylen += addrs * (sizeof(struct in6_addr) +
1460                     sizeof(u_int32_t))) > MCLBYTES)
1461                         replylen = MCLBYTES; /* XXX: will truncate pkt later */
1462                 break;
1463         case NI_QTYPE_IPV4ADDR:
1464                 /* unsupported - should respond with unknown Qtype? */
1465                 break;
1466         default:
1467                 /*
1468                  * XXX: We must return a reply with the ICMP6 code
1469                  * `unknown Qtype' in this case.  However we regard the case
1470                  * as an FQDN query for backward compatibility.
1471                  * Older versions set a random value to this field,
1472                  * so it rarely varies in the defined qtypes.
1473                  * But the mechanism is not reliable...
1474                  * maybe we should obsolete older versions.
1475                  */
1476                 qtype = NI_QTYPE_FQDN;
1477                 /* XXX will append an mbuf */
1478                 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1479                 oldfqdn++;
1480                 break;
1481         }
1482
1483         /* Allocate an mbuf to reply. */
1484         if (replylen > MCLBYTES) {
1485                 /*
1486                  * XXX: should we try to allocate more? But MCLBYTES
1487                  * is probably much larger than IPV6_MMTU...
1488                  */
1489                 goto bad;
1490         }
1491         if (replylen > MHLEN)
1492                 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1493         else
1494                 n = m_gethdr(M_NOWAIT, m->m_type);
1495         if (n == NULL) {
1496                 m_freem(m);
1497                 return (NULL);
1498         }
1499         m_move_pkthdr(n, m); /* just for recvif and FIB */
1500         n->m_pkthdr.len = n->m_len = replylen;
1501
1502         /* copy mbuf header and IPv6 + Node Information base headers */
1503         bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1504         nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1505         bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1506
1507         /* qtype dependent procedure */
1508         switch (qtype) {
1509         case NI_QTYPE_NOOP:
1510                 nni6->ni_code = ICMP6_NI_SUCCESS;
1511                 nni6->ni_flags = 0;
1512                 break;
1513         case NI_QTYPE_SUPTYPES:
1514         {
1515                 u_int32_t v;
1516                 nni6->ni_code = ICMP6_NI_SUCCESS;
1517                 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1518                 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1519                 v = (u_int32_t)htonl(0x0000000f);
1520                 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1521                 break;
1522         }
1523         case NI_QTYPE_FQDN:
1524                 nni6->ni_code = ICMP6_NI_SUCCESS;
1525                 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1526                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1527                 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1528                 fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1529                 /*
1530                  * XXX do we really have FQDN in hostname?
1531                  */
1532                 pr = curthread->td_ucred->cr_prison;
1533                 mtx_lock(&pr->pr_mtx);
1534                 n->m_next = ni6_nametodns(pr->pr_hostname,
1535                     strlen(pr->pr_hostname), oldfqdn);
1536                 mtx_unlock(&pr->pr_mtx);
1537                 if (n->m_next == NULL)
1538                         goto bad;
1539                 /* XXX we assume that n->m_next is not a chain */
1540                 if (n->m_next->m_next != NULL)
1541                         goto bad;
1542                 n->m_pkthdr.len += n->m_next->m_len;
1543                 break;
1544         case NI_QTYPE_NODEADDR:
1545         {
1546                 int lenlim, copied;
1547
1548                 nni6->ni_code = ICMP6_NI_SUCCESS;
1549                 n->m_pkthdr.len = n->m_len =
1550                     sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1551                 lenlim = M_TRAILINGSPACE(n);
1552                 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1553                 /* XXX: reset mbuf length */
1554                 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1555                     sizeof(struct icmp6_nodeinfo) + copied;
1556                 break;
1557         }
1558         default:
1559                 break;          /* XXX impossible! */
1560         }
1561
1562         nni6->ni_type = ICMP6_NI_REPLY;
1563         m_freem(m);
1564         return (n);
1565
1566   bad:
1567         m_freem(m);
1568         if (n)
1569                 m_freem(n);
1570         return (NULL);
1571 }
1572
1573 /*
1574  * make a mbuf with DNS-encoded string.  no compression support.
1575  *
1576  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1577  * treated as truncated name (two \0 at the end).  this is a wild guess.
1578  *
1579  * old - return pascal string if non-zero
1580  */
1581 static struct mbuf *
1582 ni6_nametodns(const char *name, int namelen, int old)
1583 {
1584         struct mbuf *m;
1585         char *cp, *ep;
1586         const char *p, *q;
1587         int i, len, nterm;
1588
1589         if (old)
1590                 len = namelen + 1;
1591         else
1592                 len = MCLBYTES;
1593
1594         /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1595         if (len > MLEN)
1596                 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1597         else
1598                 m = m_get(M_NOWAIT, MT_DATA);
1599         if (m == NULL)
1600                 goto fail;
1601
1602         if (old) {
1603                 m->m_len = len;
1604                 *mtod(m, char *) = namelen;
1605                 bcopy(name, mtod(m, char *) + 1, namelen);
1606                 return m;
1607         } else {
1608                 m->m_len = 0;
1609                 cp = mtod(m, char *);
1610                 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1611
1612                 /* if not certain about my name, return empty buffer */
1613                 if (namelen == 0)
1614                         return m;
1615
1616                 /*
1617                  * guess if it looks like shortened hostname, or FQDN.
1618                  * shortened hostname needs two trailing "\0".
1619                  */
1620                 i = 0;
1621                 for (p = name; p < name + namelen; p++) {
1622                         if (*p && *p == '.')
1623                                 i++;
1624                 }
1625                 if (i < 2)
1626                         nterm = 2;
1627                 else
1628                         nterm = 1;
1629
1630                 p = name;
1631                 while (cp < ep && p < name + namelen) {
1632                         i = 0;
1633                         for (q = p; q < name + namelen && *q && *q != '.'; q++)
1634                                 i++;
1635                         /* result does not fit into mbuf */
1636                         if (cp + i + 1 >= ep)
1637                                 goto fail;
1638                         /*
1639                          * DNS label length restriction, RFC1035 page 8.
1640                          * "i == 0" case is included here to avoid returning
1641                          * 0-length label on "foo..bar".
1642                          */
1643                         if (i <= 0 || i >= 64)
1644                                 goto fail;
1645                         *cp++ = i;
1646                         bcopy(p, cp, i);
1647                         cp += i;
1648                         p = q;
1649                         if (p < name + namelen && *p == '.')
1650                                 p++;
1651                 }
1652                 /* termination */
1653                 if (cp + nterm >= ep)
1654                         goto fail;
1655                 while (nterm-- > 0)
1656                         *cp++ = '\0';
1657                 m->m_len = cp - mtod(m, char *);
1658                 return m;
1659         }
1660
1661         panic("should not reach here");
1662         /* NOTREACHED */
1663
1664  fail:
1665         if (m)
1666                 m_freem(m);
1667         return NULL;
1668 }
1669
1670 /*
1671  * check if two DNS-encoded string matches.  takes care of truncated
1672  * form (with \0\0 at the end).  no compression support.
1673  * XXX upper/lowercase match (see RFC2065)
1674  */
1675 static int
1676 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1677 {
1678         const char *a0, *b0;
1679         int l;
1680
1681         /* simplest case - need validation? */
1682         if (alen == blen && bcmp(a, b, alen) == 0)
1683                 return 1;
1684
1685         a0 = a;
1686         b0 = b;
1687
1688         /* termination is mandatory */
1689         if (alen < 2 || blen < 2)
1690                 return 0;
1691         if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1692                 return 0;
1693         alen--;
1694         blen--;
1695
1696         while (a - a0 < alen && b - b0 < blen) {
1697                 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1698                         return 0;
1699
1700                 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1701                         return 0;
1702                 /* we don't support compression yet */
1703                 if (a[0] >= 64 || b[0] >= 64)
1704                         return 0;
1705
1706                 /* truncated case */
1707                 if (a[0] == 0 && a - a0 == alen - 1)
1708                         return 1;
1709                 if (b[0] == 0 && b - b0 == blen - 1)
1710                         return 1;
1711                 if (a[0] == 0 || b[0] == 0)
1712                         return 0;
1713
1714                 if (a[0] != b[0])
1715                         return 0;
1716                 l = a[0];
1717                 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1718                         return 0;
1719                 if (bcmp(a + 1, b + 1, l) != 0)
1720                         return 0;
1721
1722                 a += 1 + l;
1723                 b += 1 + l;
1724         }
1725
1726         if (a - a0 == alen && b - b0 == blen)
1727                 return 1;
1728         else
1729                 return 0;
1730 }
1731
1732 /*
1733  * calculate the number of addresses to be returned in the node info reply.
1734  */
1735 static int
1736 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1737     struct in6_addr *subj)
1738 {
1739         struct ifnet *ifp;
1740         struct in6_ifaddr *ifa6;
1741         struct ifaddr *ifa;
1742         int addrs = 0, addrsofif, iffound = 0;
1743         int niflags = ni6->ni_flags;
1744
1745         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1746                 switch (ni6->ni_code) {
1747                 case ICMP6_NI_SUBJ_IPV6:
1748                         if (subj == NULL) /* must be impossible... */
1749                                 return (0);
1750                         break;
1751                 default:
1752                         /*
1753                          * XXX: we only support IPv6 subject address for
1754                          * this Qtype.
1755                          */
1756                         return (0);
1757                 }
1758         }
1759
1760         IFNET_RLOCK_NOSLEEP();
1761         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
1762                 addrsofif = 0;
1763                 IF_ADDR_RLOCK(ifp);
1764                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1765                         if (ifa->ifa_addr->sa_family != AF_INET6)
1766                                 continue;
1767                         ifa6 = (struct in6_ifaddr *)ifa;
1768
1769                         if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1770                             IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1771                                 iffound = 1;
1772
1773                         /*
1774                          * IPv4-mapped addresses can only be returned by a
1775                          * Node Information proxy, since they represent
1776                          * addresses of IPv4-only nodes, which perforce do
1777                          * not implement this protocol.
1778                          * [icmp-name-lookups-07, Section 5.4]
1779                          * So we don't support NI_NODEADDR_FLAG_COMPAT in
1780                          * this function at this moment.
1781                          */
1782
1783                         /* What do we have to do about ::1? */
1784                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1785                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1786                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1787                                         continue;
1788                                 break;
1789                         case IPV6_ADDR_SCOPE_SITELOCAL:
1790                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1791                                         continue;
1792                                 break;
1793                         case IPV6_ADDR_SCOPE_GLOBAL:
1794                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1795                                         continue;
1796                                 break;
1797                         default:
1798                                 continue;
1799                         }
1800
1801                         /*
1802                          * check if anycast is okay.
1803                          * XXX: just experimental.  not in the spec.
1804                          */
1805                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1806                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1807                                 continue; /* we need only unicast addresses */
1808                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1809                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1810                                 continue;
1811                         }
1812                         addrsofif++; /* count the address */
1813                 }
1814                 IF_ADDR_RUNLOCK(ifp);
1815                 if (iffound) {
1816                         *ifpp = ifp;
1817                         IFNET_RUNLOCK_NOSLEEP();
1818                         return (addrsofif);
1819                 }
1820
1821                 addrs += addrsofif;
1822         }
1823         IFNET_RUNLOCK_NOSLEEP();
1824
1825         return (addrs);
1826 }
1827
1828 static int
1829 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1830     struct ifnet *ifp0, int resid)
1831 {
1832         struct ifnet *ifp;
1833         struct in6_ifaddr *ifa6;
1834         struct ifaddr *ifa;
1835         struct ifnet *ifp_dep = NULL;
1836         int copied = 0, allow_deprecated = 0;
1837         u_char *cp = (u_char *)(nni6 + 1);
1838         int niflags = ni6->ni_flags;
1839         u_int32_t ltime;
1840
1841         if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1842                 return (0);     /* needless to copy */
1843
1844         IFNET_RLOCK_NOSLEEP();
1845         ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1846   again:
1847
1848         for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1849                 IF_ADDR_RLOCK(ifp);
1850                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1851                         if (ifa->ifa_addr->sa_family != AF_INET6)
1852                                 continue;
1853                         ifa6 = (struct in6_ifaddr *)ifa;
1854
1855                         if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1856                             allow_deprecated == 0) {
1857                                 /*
1858                                  * prefererred address should be put before
1859                                  * deprecated addresses.
1860                                  */
1861
1862                                 /* record the interface for later search */
1863                                 if (ifp_dep == NULL)
1864                                         ifp_dep = ifp;
1865
1866                                 continue;
1867                         } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1868                             allow_deprecated != 0)
1869                                 continue; /* we now collect deprecated addrs */
1870
1871                         /* What do we have to do about ::1? */
1872                         switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1873                         case IPV6_ADDR_SCOPE_LINKLOCAL:
1874                                 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1875                                         continue;
1876                                 break;
1877                         case IPV6_ADDR_SCOPE_SITELOCAL:
1878                                 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1879                                         continue;
1880                                 break;
1881                         case IPV6_ADDR_SCOPE_GLOBAL:
1882                                 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1883                                         continue;
1884                                 break;
1885                         default:
1886                                 continue;
1887                         }
1888
1889                         /*
1890                          * check if anycast is okay.
1891                          * XXX: just experimental.  not in the spec.
1892                          */
1893                         if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1894                             (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1895                                 continue;
1896                         if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1897                             (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1898                                 continue;
1899                         }
1900
1901                         /* now we can copy the address */
1902                         if (resid < sizeof(struct in6_addr) +
1903                             sizeof(u_int32_t)) {
1904                                 IF_ADDR_RUNLOCK(ifp);
1905                                 /*
1906                                  * We give up much more copy.
1907                                  * Set the truncate flag and return.
1908                                  */
1909                                 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1910                                 IFNET_RUNLOCK_NOSLEEP();
1911                                 return (copied);
1912                         }
1913
1914                         /*
1915                          * Set the TTL of the address.
1916                          * The TTL value should be one of the following
1917                          * according to the specification:
1918                          *
1919                          * 1. The remaining lifetime of a DHCP lease on the
1920                          *    address, or
1921                          * 2. The remaining Valid Lifetime of a prefix from
1922                          *    which the address was derived through Stateless
1923                          *    Autoconfiguration.
1924                          *
1925                          * Note that we currently do not support stateful
1926                          * address configuration by DHCPv6, so the former
1927                          * case can't happen.
1928                          */
1929                         if (ifa6->ia6_lifetime.ia6t_expire == 0)
1930                                 ltime = ND6_INFINITE_LIFETIME;
1931                         else {
1932                                 if (ifa6->ia6_lifetime.ia6t_expire >
1933                                     time_uptime)
1934                                         ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1935                                 else
1936                                         ltime = 0;
1937                         }
1938
1939                         bcopy(&ltime, cp, sizeof(u_int32_t));
1940                         cp += sizeof(u_int32_t);
1941
1942                         /* copy the address itself */
1943                         bcopy(&ifa6->ia_addr.sin6_addr, cp,
1944                             sizeof(struct in6_addr));
1945                         in6_clearscope((struct in6_addr *)cp); /* XXX */
1946                         cp += sizeof(struct in6_addr);
1947
1948                         resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1949                         copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1950                 }
1951                 IF_ADDR_RUNLOCK(ifp);
1952                 if (ifp0)       /* we need search only on the specified IF */
1953                         break;
1954         }
1955
1956         if (allow_deprecated == 0 && ifp_dep != NULL) {
1957                 ifp = ifp_dep;
1958                 allow_deprecated = 1;
1959
1960                 goto again;
1961         }
1962
1963         IFNET_RUNLOCK_NOSLEEP();
1964
1965         return (copied);
1966 }
1967
1968 /*
1969  * XXX almost dup'ed code with rip6_input.
1970  */
1971 static int
1972 icmp6_rip6_input(struct mbuf **mp, int off)
1973 {
1974         struct mbuf *m = *mp;
1975         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1976         struct inpcb *in6p;
1977         struct inpcb *last = NULL;
1978         struct sockaddr_in6 fromsa;
1979         struct icmp6_hdr *icmp6;
1980         struct mbuf *opts = NULL;
1981
1982 #ifndef PULLDOWN_TEST
1983         /* this is assumed to be safe. */
1984         icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1985 #else
1986         IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1987         if (icmp6 == NULL) {
1988                 /* m is already reclaimed */
1989                 return (IPPROTO_DONE);
1990         }
1991 #endif
1992
1993         /*
1994          * XXX: the address may have embedded scope zone ID, which should be
1995          * hidden from applications.
1996          */
1997         bzero(&fromsa, sizeof(fromsa));
1998         fromsa.sin6_family = AF_INET6;
1999         fromsa.sin6_len = sizeof(struct sockaddr_in6);
2000         fromsa.sin6_addr = ip6->ip6_src;
2001         if (sa6_recoverscope(&fromsa)) {
2002                 m_freem(m);
2003                 return (IPPROTO_DONE);
2004         }
2005
2006         INP_INFO_RLOCK(&V_ripcbinfo);
2007         LIST_FOREACH(in6p, &V_ripcb, inp_list) {
2008                 if ((in6p->inp_vflag & INP_IPV6) == 0)
2009                         continue;
2010                 if (in6p->inp_ip_p != IPPROTO_ICMPV6)
2011                         continue;
2012                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2013                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2014                         continue;
2015                 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2016                    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2017                         continue;
2018                 INP_RLOCK(in6p);
2019                 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2020                     in6p->in6p_icmp6filt)) {
2021                         INP_RUNLOCK(in6p);
2022                         continue;
2023                 }
2024                 if (last != NULL) {
2025                         struct  mbuf *n = NULL;
2026
2027                         /*
2028                          * Recent network drivers tend to allocate a single
2029                          * mbuf cluster, rather than to make a couple of
2030                          * mbufs without clusters.  Also, since the IPv6 code
2031                          * path tries to avoid m_pullup(), it is highly
2032                          * probable that we still have an mbuf cluster here
2033                          * even though the necessary length can be stored in an
2034                          * mbuf's internal buffer.
2035                          * Meanwhile, the default size of the receive socket
2036                          * buffer for raw sockets is not so large.  This means
2037                          * the possibility of packet loss is relatively higher
2038                          * than before.  To avoid this scenario, we copy the
2039                          * received data to a separate mbuf that does not use
2040                          * a cluster, if possible.
2041                          * XXX: it is better to copy the data after stripping
2042                          * intermediate headers.
2043                          */
2044                         if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2045                             m->m_len <= MHLEN) {
2046                                 n = m_get(M_NOWAIT, m->m_type);
2047                                 if (n != NULL) {
2048                                         if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2049                                                 bcopy(m->m_data, n->m_data,
2050                                                       m->m_len);
2051                                                 n->m_len = m->m_len;
2052                                         } else {
2053                                                 m_free(n);
2054                                                 n = NULL;
2055                                         }
2056                                 }
2057                         }
2058                         if (n != NULL ||
2059                             (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2060                                 if (last->inp_flags & INP_CONTROLOPTS)
2061                                         ip6_savecontrol(last, n, &opts);
2062                                 /* strip intermediate headers */
2063                                 m_adj(n, off);
2064                                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2065                                 if (sbappendaddr_locked(
2066                                     &last->inp_socket->so_rcv,
2067                                     (struct sockaddr *)&fromsa, n, opts)
2068                                     == 0) {
2069                                         /* should notify about lost packet */
2070                                         m_freem(n);
2071                                         if (opts) {
2072                                                 m_freem(opts);
2073                                         }
2074                                         SOCKBUF_UNLOCK(
2075                                             &last->inp_socket->so_rcv);
2076                                 } else
2077                                         sorwakeup_locked(last->inp_socket);
2078                                 opts = NULL;
2079                         }
2080                         INP_RUNLOCK(last);
2081                 }
2082                 last = in6p;
2083         }
2084         INP_INFO_RUNLOCK(&V_ripcbinfo);
2085         if (last != NULL) {
2086                 if (last->inp_flags & INP_CONTROLOPTS)
2087                         ip6_savecontrol(last, m, &opts);
2088                 /* strip intermediate headers */
2089                 m_adj(m, off);
2090
2091                 /* avoid using mbuf clusters if possible (see above) */
2092                 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2093                     m->m_len <= MHLEN) {
2094                         struct mbuf *n;
2095
2096                         n = m_get(M_NOWAIT, m->m_type);
2097                         if (n != NULL) {
2098                                 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2099                                         bcopy(m->m_data, n->m_data, m->m_len);
2100                                         n->m_len = m->m_len;
2101
2102                                         m_freem(m);
2103                                         m = n;
2104                                 } else {
2105                                         m_freem(n);
2106                                         n = NULL;
2107                                 }
2108                         }
2109                 }
2110                 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2111                 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2112                     (struct sockaddr *)&fromsa, m, opts) == 0) {
2113                         m_freem(m);
2114                         if (opts)
2115                                 m_freem(opts);
2116                         SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2117                 } else
2118                         sorwakeup_locked(last->inp_socket);
2119                 INP_RUNLOCK(last);
2120         } else {
2121                 m_freem(m);
2122                 IP6STAT_DEC(ip6s_delivered);
2123         }
2124         return IPPROTO_DONE;
2125 }
2126
2127 /*
2128  * Reflect the ip6 packet back to the source.
2129  * OFF points to the icmp6 header, counted from the top of the mbuf.
2130  */
2131 void
2132 icmp6_reflect(struct mbuf *m, size_t off)
2133 {
2134         struct ip6_hdr *ip6;
2135         struct icmp6_hdr *icmp6;
2136         struct in6_ifaddr *ia = NULL;
2137         int plen;
2138         int type, code;
2139         struct ifnet *outif = NULL;
2140         struct in6_addr origdst, src, *srcp = NULL;
2141
2142         /* too short to reflect */
2143         if (off < sizeof(struct ip6_hdr)) {
2144                 nd6log((LOG_DEBUG,
2145                     "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2146                     (u_long)off, (u_long)sizeof(struct ip6_hdr),
2147                     __FILE__, __LINE__));
2148                 goto bad;
2149         }
2150
2151         /*
2152          * If there are extra headers between IPv6 and ICMPv6, strip
2153          * off that header first.
2154          */
2155 #ifdef DIAGNOSTIC
2156         if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2157                 panic("assumption failed in icmp6_reflect");
2158 #endif
2159         if (off > sizeof(struct ip6_hdr)) {
2160                 size_t l;
2161                 struct ip6_hdr nip6;
2162
2163                 l = off - sizeof(struct ip6_hdr);
2164                 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2165                 m_adj(m, l);
2166                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2167                 if (m->m_len < l) {
2168                         if ((m = m_pullup(m, l)) == NULL)
2169                                 return;
2170                 }
2171                 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2172         } else /* off == sizeof(struct ip6_hdr) */ {
2173                 size_t l;
2174                 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2175                 if (m->m_len < l) {
2176                         if ((m = m_pullup(m, l)) == NULL)
2177                                 return;
2178                 }
2179         }
2180         plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2181         ip6 = mtod(m, struct ip6_hdr *);
2182         ip6->ip6_nxt = IPPROTO_ICMPV6;
2183         icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2184         type = icmp6->icmp6_type; /* keep type for statistics */
2185         code = icmp6->icmp6_code; /* ditto. */
2186
2187         origdst = ip6->ip6_dst;
2188         /*
2189          * ip6_input() drops a packet if its src is multicast.
2190          * So, the src is never multicast.
2191          */
2192         ip6->ip6_dst = ip6->ip6_src;
2193
2194         /*
2195          * If the incoming packet was addressed directly to us (i.e. unicast),
2196          * use dst as the src for the reply.
2197          * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2198          * (for example) when we encounter an error while forwarding procedure
2199          * destined to a duplicated address of ours.
2200          * Note that ip6_getdstifaddr() may fail if we are in an error handling
2201          * procedure of an outgoing packet of our own, in which case we need
2202          * to search in the ifaddr list.
2203          */
2204         if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2205                 if ((ia = ip6_getdstifaddr(m))) {
2206                         if (!(ia->ia6_flags &
2207                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2208                                 srcp = &ia->ia_addr.sin6_addr;
2209                 } else {
2210                         struct sockaddr_in6 d;
2211
2212                         bzero(&d, sizeof(d));
2213                         d.sin6_family = AF_INET6;
2214                         d.sin6_len = sizeof(d);
2215                         d.sin6_addr = origdst;
2216                         ia = (struct in6_ifaddr *)
2217                             ifa_ifwithaddr((struct sockaddr *)&d);
2218                         if (ia &&
2219                             !(ia->ia6_flags &
2220                             (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2221                                 srcp = &ia->ia_addr.sin6_addr;
2222                         }
2223                 }
2224         }
2225
2226         if (srcp == NULL) {
2227                 int e;
2228                 struct sockaddr_in6 sin6;
2229                 struct route_in6 ro;
2230
2231                 /*
2232                  * This case matches to multicasts, our anycast, or unicasts
2233                  * that we do not own.  Select a source address based on the
2234                  * source address of the erroneous packet.
2235                  */
2236                 bzero(&sin6, sizeof(sin6));
2237                 sin6.sin6_family = AF_INET6;
2238                 sin6.sin6_len = sizeof(sin6);
2239                 sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2240
2241                 bzero(&ro, sizeof(ro));
2242                 e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2243                 if (ro.ro_rt)
2244                         RTFREE(ro.ro_rt); /* XXX: we could use this */
2245                 if (e) {
2246                         char ip6buf[INET6_ADDRSTRLEN];
2247                         nd6log((LOG_DEBUG,
2248                             "icmp6_reflect: source can't be determined: "
2249                             "dst=%s, error=%d\n",
2250                             ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2251                         goto bad;
2252                 }
2253                 srcp = &src;
2254         }
2255
2256         ip6->ip6_src = *srcp;
2257         ip6->ip6_flow = 0;
2258         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2259         ip6->ip6_vfc |= IPV6_VERSION;
2260         ip6->ip6_nxt = IPPROTO_ICMPV6;
2261         if (outif)
2262                 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2263         else if (m->m_pkthdr.rcvif) {
2264                 /* XXX: This may not be the outgoing interface */
2265                 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2266         } else
2267                 ip6->ip6_hlim = V_ip6_defhlim;
2268
2269         icmp6->icmp6_cksum = 0;
2270         icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2271             sizeof(struct ip6_hdr), plen);
2272
2273         /*
2274          * XXX option handling
2275          */
2276
2277         m->m_flags &= ~(M_BCAST|M_MCAST);
2278
2279         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2280         if (outif)
2281                 icmp6_ifoutstat_inc(outif, type, code);
2282
2283         if (ia != NULL)
2284                 ifa_free(&ia->ia_ifa);
2285         return;
2286
2287  bad:
2288         if (ia != NULL)
2289                 ifa_free(&ia->ia_ifa);
2290         m_freem(m);
2291         return;
2292 }
2293
2294 void
2295 icmp6_fasttimo(void)
2296 {
2297
2298         mld_fasttimo();
2299 }
2300
2301 void
2302 icmp6_slowtimo(void)
2303 {
2304
2305         mld_slowtimo();
2306 }
2307
2308 static const char *
2309 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2310     struct in6_addr *tgt6)
2311 {
2312         static char buf[1024];
2313         char ip6bufs[INET6_ADDRSTRLEN];
2314         char ip6bufd[INET6_ADDRSTRLEN];
2315         char ip6buft[INET6_ADDRSTRLEN];
2316         snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2317             ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2318             ip6_sprintf(ip6buft, tgt6));
2319         return buf;
2320 }
2321
2322 void
2323 icmp6_redirect_input(struct mbuf *m, int off)
2324 {
2325         struct ifnet *ifp;
2326         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2327         struct nd_redirect *nd_rd;
2328         int icmp6len = ntohs(ip6->ip6_plen);
2329         char *lladdr = NULL;
2330         int lladdrlen = 0;
2331         struct rtentry *rt = NULL;
2332         int is_router;
2333         int is_onlink;
2334         struct in6_addr src6 = ip6->ip6_src;
2335         struct in6_addr redtgt6;
2336         struct in6_addr reddst6;
2337         union nd_opts ndopts;
2338         char ip6buf[INET6_ADDRSTRLEN];
2339
2340         M_ASSERTPKTHDR(m);
2341         KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2342
2343         ifp = m->m_pkthdr.rcvif;
2344
2345         /* XXX if we are router, we don't update route by icmp6 redirect */
2346         if (V_ip6_forwarding)
2347                 goto freeit;
2348         if (!V_icmp6_rediraccept)
2349                 goto freeit;
2350
2351 #ifndef PULLDOWN_TEST
2352         IP6_EXTHDR_CHECK(m, off, icmp6len,);
2353         nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2354 #else
2355         IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2356         if (nd_rd == NULL) {
2357                 ICMP6STAT_INC(icp6s_tooshort);
2358                 return;
2359         }
2360 #endif
2361         redtgt6 = nd_rd->nd_rd_target;
2362         reddst6 = nd_rd->nd_rd_dst;
2363
2364         if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2365             in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2366                 goto freeit;
2367         }
2368
2369         /* validation */
2370         if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2371                 nd6log((LOG_ERR,
2372                     "ICMP6 redirect sent from %s rejected; "
2373                     "must be from linklocal\n",
2374                     ip6_sprintf(ip6buf, &src6)));
2375                 goto bad;
2376         }
2377         if (ip6->ip6_hlim != 255) {
2378                 nd6log((LOG_ERR,
2379                     "ICMP6 redirect sent from %s rejected; "
2380                     "hlim=%d (must be 255)\n",
2381                     ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2382                 goto bad;
2383         }
2384     {
2385         /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2386         struct sockaddr_in6 sin6;
2387         struct in6_addr *gw6;
2388
2389         bzero(&sin6, sizeof(sin6));
2390         sin6.sin6_family = AF_INET6;
2391         sin6.sin6_len = sizeof(struct sockaddr_in6);
2392         bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2393         rt = in6_rtalloc1((struct sockaddr *)&sin6, 0, 0UL, RT_DEFAULT_FIB);
2394         if (rt) {
2395                 if (rt->rt_gateway == NULL ||
2396                     rt->rt_gateway->sa_family != AF_INET6) {
2397                         RTFREE_LOCKED(rt);
2398                         nd6log((LOG_ERR,
2399                             "ICMP6 redirect rejected; no route "
2400                             "with inet6 gateway found for redirect dst: %s\n",
2401                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2402                         goto bad;
2403                 }
2404
2405                 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2406                 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2407                         RTFREE_LOCKED(rt);
2408                         nd6log((LOG_ERR,
2409                             "ICMP6 redirect rejected; "
2410                             "not equal to gw-for-src=%s (must be same): "
2411                             "%s\n",
2412                             ip6_sprintf(ip6buf, gw6),
2413                             icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2414                         goto bad;
2415                 }
2416         } else {
2417                 nd6log((LOG_ERR,
2418                     "ICMP6 redirect rejected; "
2419                     "no route found for redirect dst: %s\n",
2420                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2421                 goto bad;
2422         }
2423         RTFREE_LOCKED(rt);
2424         rt = NULL;
2425     }
2426         if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2427                 nd6log((LOG_ERR,
2428                     "ICMP6 redirect rejected; "
2429                     "redirect dst must be unicast: %s\n",
2430                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2431                 goto bad;
2432         }
2433
2434         is_router = is_onlink = 0;
2435         if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2436                 is_router = 1;  /* router case */
2437         if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2438                 is_onlink = 1;  /* on-link destination case */
2439         if (!is_router && !is_onlink) {
2440                 nd6log((LOG_ERR,
2441                     "ICMP6 redirect rejected; "
2442                     "neither router case nor onlink case: %s\n",
2443                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2444                 goto bad;
2445         }
2446         /* validation passed */
2447
2448         icmp6len -= sizeof(*nd_rd);
2449         nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2450         if (nd6_options(&ndopts) < 0) {
2451                 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2452                     __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2453                 /* nd6_options have incremented stats */
2454                 goto freeit;
2455         }
2456
2457         if (ndopts.nd_opts_tgt_lladdr) {
2458                 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2459                 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2460         }
2461
2462         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2463                 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2464                     "(if %d, icmp6 packet %d): %s\n",
2465                     __func__, ip6_sprintf(ip6buf, &redtgt6),
2466                     ifp->if_addrlen, lladdrlen - 2,
2467                     icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2468                 goto bad;
2469         }
2470
2471         /* RFC 2461 8.3 */
2472         nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2473             is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2474
2475         if (!is_onlink) {       /* better router case.  perform rtredirect. */
2476                 /* perform rtredirect */
2477                 struct sockaddr_in6 sdst;
2478                 struct sockaddr_in6 sgw;
2479                 struct sockaddr_in6 ssrc;
2480                 u_int fibnum;
2481
2482                 bzero(&sdst, sizeof(sdst));
2483                 bzero(&sgw, sizeof(sgw));
2484                 bzero(&ssrc, sizeof(ssrc));
2485                 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2486                 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2487                         sizeof(struct sockaddr_in6);
2488                 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2489                 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2490                 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2491                 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2492                         in6_rtredirect((struct sockaddr *)&sdst,
2493                             (struct sockaddr *)&sgw, (struct sockaddr *)NULL,
2494                             RTF_GATEWAY | RTF_HOST, (struct sockaddr *)&ssrc,
2495                             fibnum);
2496         }
2497         /* finally update cached route in each socket via pfctlinput */
2498     {
2499         struct sockaddr_in6 sdst;
2500
2501         bzero(&sdst, sizeof(sdst));
2502         sdst.sin6_family = AF_INET6;
2503         sdst.sin6_len = sizeof(struct sockaddr_in6);
2504         bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2505         pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2506 #ifdef IPSEC
2507         key_sa_routechange((struct sockaddr *)&sdst);
2508 #endif /* IPSEC */
2509     }
2510
2511  freeit:
2512         m_freem(m);
2513         return;
2514
2515  bad:
2516         ICMP6STAT_INC(icp6s_badredirect);
2517         m_freem(m);
2518 }
2519
2520 void
2521 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2522 {
2523         struct ifnet *ifp;      /* my outgoing interface */
2524         struct in6_addr *ifp_ll6;
2525         struct in6_addr *router_ll6;
2526         struct ip6_hdr *sip6;   /* m0 as struct ip6_hdr */
2527         struct mbuf *m = NULL;  /* newly allocated one */
2528         struct m_tag *mtag;
2529         struct ip6_hdr *ip6;    /* m as struct ip6_hdr */
2530         struct nd_redirect *nd_rd;
2531         struct llentry *ln = NULL;
2532         size_t maxlen;
2533         u_char *p;
2534         struct ifnet *outif = NULL;
2535         struct sockaddr_in6 src_sa;
2536
2537         icmp6_errcount(ND_REDIRECT, 0);
2538
2539         /* if we are not router, we don't send icmp6 redirect */
2540         if (!V_ip6_forwarding)
2541                 goto fail;
2542
2543         /* sanity check */
2544         if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2545                 goto fail;
2546
2547         /*
2548          * Address check:
2549          *  the source address must identify a neighbor, and
2550          *  the destination address must not be a multicast address
2551          *  [RFC 2461, sec 8.2]
2552          */
2553         sip6 = mtod(m0, struct ip6_hdr *);
2554         bzero(&src_sa, sizeof(src_sa));
2555         src_sa.sin6_family = AF_INET6;
2556         src_sa.sin6_len = sizeof(src_sa);
2557         src_sa.sin6_addr = sip6->ip6_src;
2558         if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2559                 goto fail;
2560         if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2561                 goto fail;      /* what should we do here? */
2562
2563         /* rate limit */
2564         if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2565                 goto fail;
2566
2567         /*
2568          * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2569          * we almost always ask for an mbuf cluster for simplicity.
2570          * (MHLEN < IPV6_MMTU is almost always true)
2571          */
2572 #if IPV6_MMTU >= MCLBYTES
2573 # error assumption failed about IPV6_MMTU and MCLBYTES
2574 #endif
2575         m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2576         if (m == NULL)
2577                 goto fail;
2578         M_SETFIB(m, rt->rt_fibnum);
2579         maxlen = M_TRAILINGSPACE(m);
2580         maxlen = min(IPV6_MMTU, maxlen);
2581         /* just for safety */
2582         if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2583             ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2584                 goto fail;
2585         }
2586
2587         {
2588                 /* get ip6 linklocal address for ifp(my outgoing interface). */
2589                 struct in6_ifaddr *ia;
2590                 if ((ia = in6ifa_ifpforlinklocal(ifp,
2591                                                  IN6_IFF_NOTREADY|
2592                                                  IN6_IFF_ANYCAST)) == NULL)
2593                         goto fail;
2594                 ifp_ll6 = &ia->ia_addr.sin6_addr;
2595                 /* XXXRW: reference released prematurely. */
2596                 ifa_free(&ia->ia_ifa);
2597         }
2598
2599         /* get ip6 linklocal address for the router. */
2600         if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2601                 struct sockaddr_in6 *sin6;
2602                 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2603                 router_ll6 = &sin6->sin6_addr;
2604                 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2605                         router_ll6 = (struct in6_addr *)NULL;
2606         } else
2607                 router_ll6 = (struct in6_addr *)NULL;
2608
2609         /* ip6 */
2610         ip6 = mtod(m, struct ip6_hdr *);
2611         ip6->ip6_flow = 0;
2612         ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2613         ip6->ip6_vfc |= IPV6_VERSION;
2614         /* ip6->ip6_plen will be set later */
2615         ip6->ip6_nxt = IPPROTO_ICMPV6;
2616         ip6->ip6_hlim = 255;
2617         /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2618         bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2619         bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2620
2621         /* ND Redirect */
2622         nd_rd = (struct nd_redirect *)(ip6 + 1);
2623         nd_rd->nd_rd_type = ND_REDIRECT;
2624         nd_rd->nd_rd_code = 0;
2625         nd_rd->nd_rd_reserved = 0;
2626         if (rt->rt_flags & RTF_GATEWAY) {
2627                 /*
2628                  * nd_rd->nd_rd_target must be a link-local address in
2629                  * better router cases.
2630                  */
2631                 if (!router_ll6)
2632                         goto fail;
2633                 bcopy(router_ll6, &nd_rd->nd_rd_target,
2634                     sizeof(nd_rd->nd_rd_target));
2635                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2636                     sizeof(nd_rd->nd_rd_dst));
2637         } else {
2638                 /* make sure redtgt == reddst */
2639                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2640                     sizeof(nd_rd->nd_rd_target));
2641                 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2642                     sizeof(nd_rd->nd_rd_dst));
2643         }
2644
2645         p = (u_char *)(nd_rd + 1);
2646
2647         if (!router_ll6)
2648                 goto nolladdropt;
2649
2650         {
2651                 /* target lladdr option */
2652                 int len;
2653                 struct nd_opt_hdr *nd_opt;
2654                 char *lladdr;
2655
2656                 IF_AFDATA_RLOCK(ifp);
2657                 ln = nd6_lookup(router_ll6, 0, ifp);
2658                 IF_AFDATA_RUNLOCK(ifp);
2659                 if (ln == NULL)
2660                         goto nolladdropt;
2661
2662                 len = sizeof(*nd_opt) + ifp->if_addrlen;
2663                 len = (len + 7) & ~7;   /* round by 8 */
2664                 /* safety check */
2665                 if (len + (p - (u_char *)ip6) > maxlen)                         
2666                         goto nolladdropt;
2667
2668                 if (ln->la_flags & LLE_VALID) {
2669                         nd_opt = (struct nd_opt_hdr *)p;
2670                         nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2671                         nd_opt->nd_opt_len = len >> 3;
2672                         lladdr = (char *)(nd_opt + 1);
2673                         bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2674                         p += len;
2675                 }
2676         }
2677 nolladdropt:
2678         if (ln != NULL)
2679                 LLE_RUNLOCK(ln);
2680                 
2681         m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2682
2683         /* just to be safe */
2684 #ifdef M_DECRYPTED      /*not openbsd*/
2685         if (m0->m_flags & M_DECRYPTED)
2686                 goto noredhdropt;
2687 #endif
2688         if (p - (u_char *)ip6 > maxlen)
2689                 goto noredhdropt;
2690
2691         {
2692                 /* redirected header option */
2693                 int len;
2694                 struct nd_opt_rd_hdr *nd_opt_rh;
2695
2696                 /*
2697                  * compute the maximum size for icmp6 redirect header option.
2698                  * XXX room for auth header?
2699                  */
2700                 len = maxlen - (p - (u_char *)ip6);
2701                 len &= ~7;
2702
2703                 /* This is just for simplicity. */
2704                 if (m0->m_pkthdr.len != m0->m_len) {
2705                         if (m0->m_next) {
2706                                 m_freem(m0->m_next);
2707                                 m0->m_next = NULL;
2708                         }
2709                         m0->m_pkthdr.len = m0->m_len;
2710                 }
2711
2712                 /*
2713                  * Redirected header option spec (RFC2461 4.6.3) talks nothing
2714                  * about padding/truncate rule for the original IP packet.
2715                  * From the discussion on IPv6imp in Feb 1999,
2716                  * the consensus was:
2717                  * - "attach as much as possible" is the goal
2718                  * - pad if not aligned (original size can be guessed by
2719                  *   original ip6 header)
2720                  * Following code adds the padding if it is simple enough,
2721                  * and truncates if not.
2722                  */
2723                 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2724                         panic("assumption failed in %s:%d", __FILE__,
2725                             __LINE__);
2726
2727                 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2728                         /* not enough room, truncate */
2729                         m0->m_pkthdr.len = m0->m_len = len -
2730                             sizeof(*nd_opt_rh);
2731                 } else {
2732                         /* enough room, pad or truncate */
2733                         size_t extra;
2734
2735                         extra = m0->m_pkthdr.len % 8;
2736                         if (extra) {
2737                                 /* pad if easy enough, truncate if not */
2738                                 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2739                                         /* pad */
2740                                         m0->m_len += (8 - extra);
2741                                         m0->m_pkthdr.len += (8 - extra);
2742                                 } else {
2743                                         /* truncate */
2744                                         m0->m_pkthdr.len -= extra;
2745                                         m0->m_len -= extra;
2746                                 }
2747                         }
2748                         len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2749                         m0->m_pkthdr.len = m0->m_len = len -
2750                             sizeof(*nd_opt_rh);
2751                 }
2752
2753                 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2754                 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2755                 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2756                 nd_opt_rh->nd_opt_rh_len = len >> 3;
2757                 p += sizeof(*nd_opt_rh);
2758                 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2759
2760                 /* connect m0 to m */
2761                 m_tag_delete_chain(m0, NULL);
2762                 m0->m_flags &= ~M_PKTHDR;
2763                 m->m_next = m0;
2764                 m->m_pkthdr.len = m->m_len + m0->m_len;
2765                 m0 = NULL;
2766         }
2767 noredhdropt:;
2768         if (m0) {
2769                 m_freem(m0);
2770                 m0 = NULL;
2771         }
2772
2773         /* XXX: clear embedded link IDs in the inner header */
2774         in6_clearscope(&sip6->ip6_src);
2775         in6_clearscope(&sip6->ip6_dst);
2776         in6_clearscope(&nd_rd->nd_rd_target);
2777         in6_clearscope(&nd_rd->nd_rd_dst);
2778
2779         ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2780
2781         nd_rd->nd_rd_cksum = 0;
2782         nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2783             sizeof(*ip6), ntohs(ip6->ip6_plen));
2784
2785         if (send_sendso_input_hook != NULL) {
2786                 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2787                         M_NOWAIT);
2788                 if (mtag == NULL)
2789                         goto fail;
2790                 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2791                 m_tag_prepend(m, mtag);
2792         }
2793
2794         /* send the packet to outside... */
2795         ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2796         if (outif) {
2797                 icmp6_ifstat_inc(outif, ifs6_out_msg);
2798                 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2799         }
2800         ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2801
2802         return;
2803
2804 fail:
2805         if (m)
2806                 m_freem(m);
2807         if (m0)
2808                 m_freem(m0);
2809 }
2810
2811 /*
2812  * ICMPv6 socket option processing.
2813  */
2814 int
2815 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2816 {
2817         int error = 0;
2818         int optlen;
2819         struct inpcb *inp = sotoinpcb(so);
2820         int level, op, optname;
2821
2822         if (sopt) {
2823                 level = sopt->sopt_level;
2824                 op = sopt->sopt_dir;
2825                 optname = sopt->sopt_name;
2826                 optlen = sopt->sopt_valsize;
2827         } else
2828                 level = op = optname = optlen = 0;
2829
2830         if (level != IPPROTO_ICMPV6) {
2831                 return EINVAL;
2832         }
2833
2834         switch (op) {
2835         case PRCO_SETOPT:
2836                 switch (optname) {
2837                 case ICMP6_FILTER:
2838                     {
2839                         struct icmp6_filter ic6f;
2840
2841                         if (optlen != sizeof(ic6f)) {
2842                                 error = EMSGSIZE;
2843                                 break;
2844                         }
2845                         error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2846                         if (error == 0) {
2847                                 INP_WLOCK(inp);
2848                                 *inp->in6p_icmp6filt = ic6f;
2849                                 INP_WUNLOCK(inp);
2850                         }
2851                         break;
2852                     }
2853
2854                 default:
2855                         error = ENOPROTOOPT;
2856                         break;
2857                 }
2858                 break;
2859
2860         case PRCO_GETOPT:
2861                 switch (optname) {
2862                 case ICMP6_FILTER:
2863                     {
2864                         struct icmp6_filter ic6f;
2865
2866                         INP_RLOCK(inp);
2867                         ic6f = *inp->in6p_icmp6filt;
2868                         INP_RUNLOCK(inp);
2869                         error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2870                         break;
2871                     }
2872
2873                 default:
2874                         error = ENOPROTOOPT;
2875                         break;
2876                 }
2877                 break;
2878         }
2879
2880         return (error);
2881 }
2882
2883 /*
2884  * Perform rate limit check.
2885  * Returns 0 if it is okay to send the icmp6 packet.
2886  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2887  * limitation.
2888  *
2889  * XXX per-destination/type check necessary?
2890  *
2891  * dst - not used at this moment
2892  * type - not used at this moment
2893  * code - not used at this moment
2894  */
2895 static int
2896 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2897     const int code)
2898 {
2899         int ret;
2900
2901         ret = 0;        /* okay to send */
2902
2903         /* PPS limit */
2904         if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2905             V_icmp6errppslim)) {
2906                 /* The packet is subject to rate limit */
2907                 ret++;
2908         }
2909
2910         return ret;
2911 }